Merge "Use full-pixel MV in mvsadcost calculation"
diff --git a/build/make/obj_int_extract.c b/build/make/obj_int_extract.c
index c46d9d5..04e14a6 100644
--- a/build/make/obj_int_extract.c
+++ b/build/make/obj_int_extract.c
@@ -9,25 +9,13 @@
*/
+#include <stdarg.h>
#include <stdio.h>
#include <stdlib.h>
+#include <string.h>
#include "vpx_config.h"
-
-#if defined(_MSC_VER) || defined(__MINGW32__)
-#include <io.h>
-#include <share.h>
#include "vpx/vpx_integer.h"
-#else
-#include <stdint.h>
-#include <unistd.h>
-#endif
-
-#include <string.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <stdarg.h>
typedef enum
{
@@ -47,7 +35,6 @@
}
#if defined(__GNUC__) && __GNUC__
-
#if defined(__MACH__)
#include <mach-o/loader.h>
@@ -225,73 +212,6 @@
}
-int main(int argc, char **argv)
-{
- int fd;
- char *f;
- struct stat stat_buf;
- uint8_t *file_buf;
- int res;
-
- if (argc < 2 || argc > 3)
- {
- fprintf(stderr, "Usage: %s [output format] <obj file>\n\n", argv[0]);
- fprintf(stderr, " <obj file>\tMachO format object file to parse\n");
- fprintf(stderr, "Output Formats:\n");
- fprintf(stderr, " gas - compatible with GNU assembler\n");
- fprintf(stderr, " rvds - compatible with armasm\n");
- goto bail;
- }
-
- f = argv[2];
-
- if (!((!strcmp(argv[1], "rvds")) || (!strcmp(argv[1], "gas"))))
- f = argv[1];
-
- fd = open(f, O_RDONLY);
-
- if (fd < 0)
- {
- perror("Unable to open file");
- goto bail;
- }
-
- if (fstat(fd, &stat_buf))
- {
- perror("stat");
- goto bail;
- }
-
- file_buf = malloc(stat_buf.st_size);
-
- if (!file_buf)
- {
- perror("malloc");
- goto bail;
- }
-
- if (read(fd, file_buf, stat_buf.st_size) != stat_buf.st_size)
- {
- perror("read");
- goto bail;
- }
-
- if (close(fd))
- {
- perror("close");
- goto bail;
- }
-
- res = parse_macho(file_buf, stat_buf.st_size);
- free(file_buf);
-
- if (!res)
- return EXIT_SUCCESS;
-
-bail:
- return EXIT_FAILURE;
-}
-
#elif defined(__ELF__)
#include "elf.h"
@@ -740,96 +660,24 @@
return 1;
}
-int main(int argc, char **argv)
-{
- int fd;
- output_fmt_t mode;
- char *f;
- struct stat stat_buf;
- uint8_t *file_buf;
- int res;
-
- if (argc < 2 || argc > 3)
- {
- fprintf(stderr, "Usage: %s [output format] <obj file>\n\n", argv[0]);
- fprintf(stderr, " <obj file>\tELF format object file to parse\n");
- fprintf(stderr, "Output Formats:\n");
- fprintf(stderr, " gas - compatible with GNU assembler\n");
- fprintf(stderr, " rvds - compatible with armasm\n");
- goto bail;
- }
-
- f = argv[2];
-
- if (!strcmp(argv[1], "rvds"))
- mode = OUTPUT_FMT_RVDS;
- else if (!strcmp(argv[1], "gas"))
- mode = OUTPUT_FMT_GAS;
- else
- f = argv[1];
-
-
- fd = open(f, O_RDONLY);
-
- if (fd < 0)
- {
- perror("Unable to open file");
- goto bail;
- }
-
- if (fstat(fd, &stat_buf))
- {
- perror("stat");
- goto bail;
- }
-
- file_buf = malloc(stat_buf.st_size);
-
- if (!file_buf)
- {
- perror("malloc");
- goto bail;
- }
-
- if (read(fd, file_buf, stat_buf.st_size) != stat_buf.st_size)
- {
- perror("read");
- goto bail;
- }
-
- if (close(fd))
- {
- perror("close");
- goto bail;
- }
-
- res = parse_elf(file_buf, stat_buf.st_size, mode);
- free(file_buf);
-
- if (!res)
- return EXIT_SUCCESS;
-
-bail:
- return EXIT_FAILURE;
-}
#endif
-#endif
+#endif /* defined(__GNUC__) && __GNUC__ */
-#if defined(_MSC_VER) || defined(__MINGW32__)
+#if defined(_MSC_VER) || defined(__MINGW32__) || defined(__CYGWIN__)
/* See "Microsoft Portable Executable and Common Object File Format Specification"
for reference.
*/
#define get_le32(x) ((*(x)) | (*(x+1)) << 8 |(*(x+2)) << 16 | (*(x+3)) << 24 )
#define get_le16(x) ((*(x)) | (*(x+1)) << 8)
-int parse_coff(unsigned __int8 *buf, size_t sz)
+int parse_coff(uint8_t *buf, size_t sz)
{
unsigned int nsections, symtab_ptr, symtab_sz, strtab_ptr;
unsigned int sectionrawdata_ptr;
unsigned int i;
- unsigned __int8 *ptr;
- unsigned __int32 symoffset;
+ uint8_t *ptr;
+ uint32_t symoffset;
char **sectionlist; //this array holds all section names in their correct order.
//it is used to check if the symbol is in .bss or .data section.
@@ -907,7 +755,7 @@
for (i = 0; i < symtab_sz; i++)
{
- __int16 section = get_le16(ptr + 12); //section number
+ int16_t section = get_le16(ptr + 12); //section number
if (section > 0 && ptr[16] == 2)
{
@@ -978,20 +826,21 @@
return 1;
}
+#endif /* defined(_MSC_VER) || defined(__MINGW32__) || defined(__CYGWIN__) */
int main(int argc, char **argv)
{
- int fd;
- output_fmt_t mode;
+ output_fmt_t mode = OUTPUT_FMT_PLAIN;
const char *f;
- struct _stat stat_buf;
- unsigned __int8 *file_buf;
+ uint8_t *file_buf;
int res;
+ FILE *fp;
+ long int file_size;
if (argc < 2 || argc > 3)
{
fprintf(stderr, "Usage: %s [output format] <obj file>\n\n", argv[0]);
- fprintf(stderr, " <obj file>\tELF format object file to parse\n");
+ fprintf(stderr, " <obj file>\tobject file to parse\n");
fprintf(stderr, "Output Formats:\n");
fprintf(stderr, " gas - compatible with GNU assembler\n");
fprintf(stderr, " rvds - compatible with armasm\n");
@@ -1007,15 +856,22 @@
else
f = argv[1];
- fd = _sopen(f, _O_BINARY, _SH_DENYNO, _S_IREAD | _S_IWRITE);
+ fp = fopen(f, "rb");
- if (_fstat(fd, &stat_buf))
+ if (!fp)
+ {
+ perror("Unable to open file");
+ goto bail;
+ }
+
+ if (fseek(fp, 0, SEEK_END))
{
perror("stat");
goto bail;
}
- file_buf = malloc(stat_buf.st_size);
+ file_size = ftell(fp);
+ file_buf = malloc(file_size);
if (!file_buf)
{
@@ -1023,19 +879,30 @@
goto bail;
}
- if (_read(fd, file_buf, stat_buf.st_size) != stat_buf.st_size)
+ rewind(fp);
+
+ if (fread(file_buf, sizeof(char), file_size, fp) != file_size)
{
perror("read");
goto bail;
}
- if (_close(fd))
+ if (fclose(fp))
{
perror("close");
goto bail;
}
- res = parse_coff(file_buf, stat_buf.st_size);
+#if defined(__GNUC__) && __GNUC__
+#if defined(__MACH__)
+ res = parse_macho(file_buf, file_size);
+#elif defined(__ELF__)
+ res = parse_elf(file_buf, file_size, mode);
+#endif
+#endif
+#if defined(_MSC_VER) || defined(__MINGW32__) || defined(__CYGWIN__)
+ res = parse_coff(file_buf, file_size);
+#endif
free(file_buf);
@@ -1045,4 +912,3 @@
bail:
return EXIT_FAILURE;
}
-#endif
diff --git a/configure b/configure
index c177865..ab3936d 100755
--- a/configure
+++ b/configure
@@ -377,6 +377,7 @@
if [ -f "${source_path}/build/make/version.sh" ]; then
local ver=`"$source_path/build/make/version.sh" --bare $source_path`
DIST_DIR="${DIST_DIR}-${ver}"
+ VERSION_STRING=${ver}
ver=${ver%%-*}
VERSION_PATCH=${ver##*.}
ver=${ver%.*}
@@ -385,6 +386,8 @@
VERSION_MAJOR=${ver%.*}
fi
enabled child || cat <<EOF >> config.mk
+
+PREFIX=${prefix}
ifeq (\$(MAKECMDGOALS),dist)
DIST_DIR?=${DIST_DIR}
else
@@ -392,6 +395,8 @@
endif
LIBSUBDIR=${libdir##${prefix}/}
+VERSION_STRING=${VERSION_STRING}
+
VERSION_MAJOR=${VERSION_MAJOR}
VERSION_MINOR=${VERSION_MINOR}
VERSION_PATCH=${VERSION_PATCH}
diff --git a/libs.mk b/libs.mk
index 6a5dc18..2cb7f49 100644
--- a/libs.mk
+++ b/libs.mk
@@ -204,6 +204,26 @@
INSTALL-LIBS-$(CONFIG_SHARED) += $(LIBVPX_SO_SYMLINKS)
INSTALL-LIBS-$(CONFIG_SHARED) += $(LIBSUBDIR)/$(LIBVPX_SO)
+
+LIBS-$(BUILD_LIBVPX) += vpx.pc
+vpx.pc: config.mk libs.mk
+ @echo " [CREATE] $@"
+ $(qexec)echo '# pkg-config file from libvpx $(VERSION_STRING)' > $@
+ $(qexec)echo 'prefix=$(PREFIX)' >> $@
+ $(qexec)echo 'exec_prefix=$${prefix}' >> $@
+ $(qexec)echo 'libdir=$${prefix}/lib' >> $@
+ $(qexec)echo 'includedir=$${prefix}/include' >> $@
+ $(qexec)echo '' >> $@
+ $(qexec)echo 'Name: vpx' >> $@
+ $(qexec)echo 'Description: WebM Project VPx codec implementation' >> $@
+ $(qexec)echo 'Version: $(VERSION_MAJOR).$(VERSION_MINOR).$(VERSION_PATCH)' >> $@
+ $(qexec)echo 'Requires:' >> $@
+ $(qexec)echo 'Conflicts:' >> $@
+ $(qexec)echo 'Libs: -L$${libdir} -lvpx' >> $@
+ $(qexec)echo 'Cflags: -I$${includedir}' >> $@
+INSTALL-LIBS-yes += $(LIBSUBDIR)/pkgconfig/vpx.pc
+INSTALL_MAPS += $(LIBSUBDIR)/pkgconfig/%.pc %.pc
+CLEAN-OBJS += vpx.pc
endif
LIBS-$(LIPO_LIBVPX) += libvpx.a
diff --git a/vp8/encoder/arm/arm_csystemdependent.c b/vp8/encoder/arm/arm_csystemdependent.c
index d4a88dc..f148b89 100644
--- a/vp8/encoder/arm/arm_csystemdependent.c
+++ b/vp8/encoder/arm/arm_csystemdependent.c
@@ -66,10 +66,10 @@
/*cpi->rtcd.encodemb.berr = vp8_block_error_c;
cpi->rtcd.encodemb.mberr = vp8_mbblock_error_c;
- cpi->rtcd.encodemb.mbuverr = vp8_mbuverror_c;
- cpi->rtcd.encodemb.subb = vp8_subtract_b_c;
- cpi->rtcd.encodemb.submby = vp8_subtract_mby_c;
- cpi->rtcd.encodemb.submbuv = vp8_subtract_mbuv_c;*/
+ cpi->rtcd.encodemb.mbuverr = vp8_mbuverror_c;*/
+ cpi->rtcd.encodemb.subb = vp8_subtract_b_armv6;
+ cpi->rtcd.encodemb.submby = vp8_subtract_mby_armv6;
+ cpi->rtcd.encodemb.submbuv = vp8_subtract_mbuv_armv6;
/*cpi->rtcd.quantize.quantb = vp8_regular_quantize_b;*/
cpi->rtcd.quantize.fastquantb = vp8_fast_quantize_b_armv6;
diff --git a/vp8/encoder/arm/armv6/vp8_subtract_armv6.asm b/vp8/encoder/arm/armv6/vp8_subtract_armv6.asm
new file mode 100644
index 0000000..0ca7438
--- /dev/null
+++ b/vp8/encoder/arm/armv6/vp8_subtract_armv6.asm
@@ -0,0 +1,265 @@
+;
+; Copyright (c) 2011 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+ EXPORT |vp8_subtract_mby_armv6|
+ EXPORT |vp8_subtract_mbuv_armv6|
+ EXPORT |vp8_subtract_b_armv6|
+
+ INCLUDE asm_enc_offsets.asm
+
+ ARM
+ REQUIRE8
+ PRESERVE8
+
+ AREA ||.text||, CODE, READONLY, ALIGN=2
+
+; r0 BLOCK *be
+; r1 BLOCKD *bd
+; r2 int pitch
+|vp8_subtract_b_armv6| PROC
+
+ stmfd sp!, {r4-r9}
+
+ ldr r4, [r0, #vp8_block_base_src]
+ ldr r5, [r0, #vp8_block_src]
+ ldr r6, [r0, #vp8_block_src_diff]
+
+ ldr r3, [r4]
+ ldr r7, [r0, #vp8_block_src_stride]
+ add r3, r3, r5 ; src = *base_src + src
+ ldr r8, [r1, #vp8_blockd_predictor]
+
+ mov r9, #4 ; loop count
+
+loop_block
+
+ ldr r0, [r3], r7 ; src
+ ldr r1, [r8], r2 ; pred
+
+ uxtb16 r4, r0 ; [s2 | s0]
+ uxtb16 r5, r1 ; [p2 | p0]
+ uxtb16 r0, r0, ror #8 ; [s3 | s1]
+ uxtb16 r1, r1, ror #8 ; [p3 | p1]
+
+ usub16 r4, r4, r5 ; [d2 | d0]
+ usub16 r5, r0, r1 ; [d3 | d1]
+
+ subs r9, r9, #1 ; decrement loop counter
+
+ pkhbt r0, r4, r5, lsl #16 ; [d1 | d0]
+ pkhtb r1, r5, r4, asr #16 ; [d3 | d2]
+
+ str r0, [r6, #0] ; diff
+ str r1, [r6, #4] ; diff
+
+ add r6, r6, r2, lsl #1 ; update diff pointer
+ bne loop_block
+
+ ldmfd sp!, {r4-r9}
+ mov pc, lr
+
+ ENDP
+
+
+; r0 short *diff
+; r1 unsigned char *usrc
+; r2 unsigned char *vsrc
+; r3 unsigned char *pred
+; stack int stride
+|vp8_subtract_mbuv_armv6| PROC
+
+ stmfd sp!, {r4-r12, lr}
+
+ add r0, r0, #512 ; set *diff point to Cb
+ add r3, r3, #256 ; set *pred point to Cb
+
+ mov r4, #8 ; loop count
+ ldr r5, [sp, #40] ; stride
+
+ ; Subtract U block
+loop_u
+ ldr r6, [r1] ; src (A)
+ ldr r7, [r3], #4 ; pred (A)
+
+ uxtb16 r8, r6 ; [s2 | s0] (A)
+ uxtb16 r9, r7 ; [p2 | p0] (A)
+ uxtb16 r10, r6, ror #8 ; [s3 | s1] (A)
+ uxtb16 r11, r7, ror #8 ; [p3 | p1] (A)
+
+ usub16 r6, r8, r9 ; [d2 | d0] (A)
+ usub16 r7, r10, r11 ; [d3 | d1] (A)
+
+ ldr r10, [r1, #4] ; src (B)
+ ldr r11, [r3], #4 ; pred (B)
+
+ pkhbt r8, r6, r7, lsl #16 ; [d1 | d0] (A)
+ pkhtb r9, r7, r6, asr #16 ; [d3 | d2] (A)
+
+ str r8, [r0], #4 ; diff (A)
+ uxtb16 r8, r10 ; [s2 | s0] (B)
+ str r9, [r0], #4 ; diff (A)
+
+ uxtb16 r9, r11 ; [p2 | p0] (B)
+ uxtb16 r10, r10, ror #8 ; [s3 | s1] (B)
+ uxtb16 r11, r11, ror #8 ; [p3 | p1] (B)
+
+ usub16 r6, r8, r9 ; [d2 | d0] (B)
+ usub16 r7, r10, r11 ; [d3 | d1] (B)
+
+ add r1, r1, r5 ; update usrc pointer
+
+ pkhbt r8, r6, r7, lsl #16 ; [d1 | d0] (B)
+ pkhtb r9, r7, r6, asr #16 ; [d3 | d2] (B)
+
+ str r8, [r0], #4 ; diff (B)
+ subs r4, r4, #1 ; update loop counter
+ str r9, [r0], #4 ; diff (B)
+
+ bne loop_u
+
+ mov r4, #8 ; loop count
+
+ ; Subtract V block
+loop_v
+ ldr r6, [r2] ; src (A)
+ ldr r7, [r3], #4 ; pred (A)
+
+ uxtb16 r8, r6 ; [s2 | s0] (A)
+ uxtb16 r9, r7 ; [p2 | p0] (A)
+ uxtb16 r10, r6, ror #8 ; [s3 | s1] (A)
+ uxtb16 r11, r7, ror #8 ; [p3 | p1] (A)
+
+ usub16 r6, r8, r9 ; [d2 | d0] (A)
+ usub16 r7, r10, r11 ; [d3 | d1] (A)
+
+ ldr r10, [r2, #4] ; src (B)
+ ldr r11, [r3], #4 ; pred (B)
+
+ pkhbt r8, r6, r7, lsl #16 ; [d1 | d0] (A)
+ pkhtb r9, r7, r6, asr #16 ; [d3 | d2] (A)
+
+ str r8, [r0], #4 ; diff (A)
+ uxtb16 r8, r10 ; [s2 | s0] (B)
+ str r9, [r0], #4 ; diff (A)
+
+ uxtb16 r9, r11 ; [p2 | p0] (B)
+ uxtb16 r10, r10, ror #8 ; [s3 | s1] (B)
+ uxtb16 r11, r11, ror #8 ; [p3 | p1] (B)
+
+ usub16 r6, r8, r9 ; [d2 | d0] (B)
+ usub16 r7, r10, r11 ; [d3 | d1] (B)
+
+ add r2, r2, r5 ; update vsrc pointer
+
+ pkhbt r8, r6, r7, lsl #16 ; [d1 | d0] (B)
+ pkhtb r9, r7, r6, asr #16 ; [d3 | d2] (B)
+
+ str r8, [r0], #4 ; diff (B)
+ subs r4, r4, #1 ; update loop counter
+ str r9, [r0], #4 ; diff (B)
+
+ bne loop_v
+
+ ldmfd sp!, {r4-r12, pc}
+
+ ENDP
+
+
+; r0 short *diff
+; r1 unsigned char *src
+; r2 unsigned char *pred
+; r3 int stride
+|vp8_subtract_mby_armv6| PROC
+
+ stmfd sp!, {r4-r11}
+
+ mov r4, #16
+loop
+ ldr r6, [r1] ; src (A)
+ ldr r7, [r2], #4 ; pred (A)
+
+ uxtb16 r8, r6 ; [s2 | s0] (A)
+ uxtb16 r9, r7 ; [p2 | p0] (A)
+ uxtb16 r10, r6, ror #8 ; [s3 | s1] (A)
+ uxtb16 r11, r7, ror #8 ; [p3 | p1] (A)
+
+ usub16 r6, r8, r9 ; [d2 | d0] (A)
+ usub16 r7, r10, r11 ; [d3 | d1] (A)
+
+ ldr r10, [r1, #4] ; src (B)
+ ldr r11, [r2], #4 ; pred (B)
+
+ pkhbt r8, r6, r7, lsl #16 ; [d1 | d0] (A)
+ pkhtb r9, r7, r6, asr #16 ; [d3 | d2] (A)
+
+ str r8, [r0], #4 ; diff (A)
+ uxtb16 r8, r10 ; [s2 | s0] (B)
+ str r9, [r0], #4 ; diff (A)
+
+ uxtb16 r9, r11 ; [p2 | p0] (B)
+ uxtb16 r10, r10, ror #8 ; [s3 | s1] (B)
+ uxtb16 r11, r11, ror #8 ; [p3 | p1] (B)
+
+ usub16 r6, r8, r9 ; [d2 | d0] (B)
+ usub16 r7, r10, r11 ; [d3 | d1] (B)
+
+ ldr r10, [r1, #8] ; src (C)
+ ldr r11, [r2], #4 ; pred (C)
+
+ pkhbt r8, r6, r7, lsl #16 ; [d1 | d0] (B)
+ pkhtb r9, r7, r6, asr #16 ; [d3 | d2] (B)
+
+ str r8, [r0], #4 ; diff (B)
+ uxtb16 r8, r10 ; [s2 | s0] (C)
+ str r9, [r0], #4 ; diff (B)
+
+ uxtb16 r9, r11 ; [p2 | p0] (C)
+ uxtb16 r10, r10, ror #8 ; [s3 | s1] (C)
+ uxtb16 r11, r11, ror #8 ; [p3 | p1] (C)
+
+ usub16 r6, r8, r9 ; [d2 | d0] (C)
+ usub16 r7, r10, r11 ; [d3 | d1] (C)
+
+ ldr r10, [r1, #12] ; src (D)
+ ldr r11, [r2], #4 ; pred (D)
+
+ pkhbt r8, r6, r7, lsl #16 ; [d1 | d0] (C)
+ pkhtb r9, r7, r6, asr #16 ; [d3 | d2] (C)
+
+ str r8, [r0], #4 ; diff (C)
+ uxtb16 r8, r10 ; [s2 | s0] (D)
+ str r9, [r0], #4 ; diff (C)
+
+ uxtb16 r9, r11 ; [p2 | p0] (D)
+ uxtb16 r10, r10, ror #8 ; [s3 | s1] (D)
+ uxtb16 r11, r11, ror #8 ; [p3 | p1] (D)
+
+ usub16 r6, r8, r9 ; [d2 | d0] (D)
+ usub16 r7, r10, r11 ; [d3 | d1] (D)
+
+ add r1, r1, r3 ; update src pointer
+
+ pkhbt r8, r6, r7, lsl #16 ; [d1 | d0] (D)
+ pkhtb r9, r7, r6, asr #16 ; [d3 | d2] (D)
+
+ str r8, [r0], #4 ; diff (D)
+ subs r4, r4, #1 ; update loop counter
+ str r9, [r0], #4 ; diff (D)
+
+ bne loop
+
+ ldmfd sp!, {r4-r11}
+ mov pc, lr
+
+ ENDP
+
+ END
+
diff --git a/vp8/encoder/arm/armv6/vp8_variance16x16_armv6.asm b/vp8/encoder/arm/armv6/vp8_variance16x16_armv6.asm
index 8d7258a..9883763 100644
--- a/vp8/encoder/arm/armv6/vp8_variance16x16_armv6.asm
+++ b/vp8/encoder/arm/armv6/vp8_variance16x16_armv6.asm
@@ -25,14 +25,14 @@
|vp8_variance16x16_armv6| PROC
stmfd sp!, {r4-r12, lr}
- mov r12, #16 ; set loop counter to 16 (=block height)
mov r8, #0 ; initialize sum = 0
mov r11, #0 ; initialize sse = 0
+ mov r12, #16 ; set loop counter to 16 (=block height)
loop
; 1st 4 pixels
- ldr r4, [r0, #0x0] ; load 4 src pixels
- ldr r5, [r2, #0x0] ; load 4 ref pixels
+ ldr r4, [r0, #0] ; load 4 src pixels
+ ldr r5, [r2, #0] ; load 4 ref pixels
mov lr, #0 ; constant zero
@@ -55,8 +55,8 @@
smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
; 2nd 4 pixels
- ldr r4, [r0, #0x4] ; load 4 src pixels
- ldr r5, [r2, #0x4] ; load 4 ref pixels
+ ldr r4, [r0, #4] ; load 4 src pixels
+ ldr r5, [r2, #4] ; load 4 ref pixels
smlad r11, r10, r10, r11 ; dual signed multiply, add and accumulate (2)
usub8 r6, r4, r5 ; calculate difference
@@ -79,8 +79,8 @@
smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
; 3rd 4 pixels
- ldr r4, [r0, #0x8] ; load 4 src pixels
- ldr r5, [r2, #0x8] ; load 4 ref pixels
+ ldr r4, [r0, #8] ; load 4 src pixels
+ ldr r5, [r2, #8] ; load 4 ref pixels
smlad r11, r10, r10, r11 ; dual signed multiply, add and accumulate (2)
usub8 r6, r4, r5 ; calculate difference
@@ -103,8 +103,8 @@
smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
; 4th 4 pixels
- ldr r4, [r0, #0xc] ; load 4 src pixels
- ldr r5, [r2, #0xc] ; load 4 ref pixels
+ ldr r4, [r0, #12] ; load 4 src pixels
+ ldr r5, [r2, #12] ; load 4 ref pixels
smlad r11, r10, r10, r11 ; dual signed multiply, add and accumulate (2)
usub8 r6, r4, r5 ; calculate difference
@@ -135,13 +135,14 @@
bne loop
; return stuff
- ldr r6, [sp, #0x28] ; get address of sse
+ ldr r6, [sp, #40] ; get address of sse
mul r0, r8, r8 ; sum * sum
str r11, [r6] ; store sse
- sub r0, r11, r0, ASR #8 ; return (sse - ((sum * sum) >> 8))
+ sub r0, r11, r0, asr #8 ; return (sse - ((sum * sum) >> 8))
ldmfd sp!, {r4-r12, pc}
ENDP
END
+
diff --git a/vp8/encoder/arm/armv6/vp8_variance_halfpixvar16x16_h_armv6.asm b/vp8/encoder/arm/armv6/vp8_variance_halfpixvar16x16_h_armv6.asm
new file mode 100644
index 0000000..2350f3e
--- /dev/null
+++ b/vp8/encoder/arm/armv6/vp8_variance_halfpixvar16x16_h_armv6.asm
@@ -0,0 +1,176 @@
+;
+; Copyright (c) 2011 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+ EXPORT |vp8_variance_halfpixvar16x16_h_armv6|
+
+ ARM
+ REQUIRE8
+ PRESERVE8
+
+ AREA ||.text||, CODE, READONLY, ALIGN=2
+
+; r0 unsigned char *src_ptr
+; r1 int source_stride
+; r2 unsigned char *ref_ptr
+; r3 int recon_stride
+; stack unsigned int *sse
+|vp8_variance_halfpixvar16x16_h_armv6| PROC
+
+ stmfd sp!, {r4-r12, lr}
+ mov r8, #0 ; initialize sum = 0
+ ldr r10, c80808080
+ mov r11, #0 ; initialize sse = 0
+ mov r12, #16 ; set loop counter to 16 (=block height)
+ mov lr, #0 ; constant zero
+loop
+ ; 1st 4 pixels
+ ldr r4, [r0, #0] ; load 4 src pixels
+ ldr r6, [r0, #1] ; load 4 src pixels with 1 byte offset
+ ldr r5, [r2, #0] ; load 4 ref pixels
+
+ ; bilinear interpolation
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+
+ usub8 r6, r4, r5 ; calculate difference
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+ ; calculate total sum
+ adds r8, r8, r4 ; add positive differences to sum
+ subs r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 2nd 4 pixels
+ ldr r4, [r0, #4] ; load 4 src pixels
+ ldr r6, [r0, #5] ; load 4 src pixels with 1 byte offset
+ ldr r5, [r2, #4] ; load 4 ref pixels
+
+ ; bilinear interpolation
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+ usub8 r6, r4, r5 ; calculate difference
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 3rd 4 pixels
+ ldr r4, [r0, #8] ; load 4 src pixels
+ ldr r6, [r0, #9] ; load 4 src pixels with 1 byte offset
+ ldr r5, [r2, #8] ; load 4 ref pixels
+
+ ; bilinear interpolation
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+ usub8 r6, r4, r5 ; calculate difference
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 4th 4 pixels
+ ldr r4, [r0, #12] ; load 4 src pixels
+ ldr r6, [r0, #13] ; load 4 src pixels with 1 byte offset
+ ldr r5, [r2, #12] ; load 4 ref pixels
+
+ ; bilinear interpolation
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+ usub8 r6, r4, r5 ; calculate difference
+ add r0, r0, r1 ; set src_ptr to next row
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ add r2, r2, r3 ; set dst_ptr to next row
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+ subs r12, r12, #1
+
+ bne loop
+
+ ; return stuff
+ ldr r6, [sp, #40] ; get address of sse
+ mul r0, r8, r8 ; sum * sum
+ str r11, [r6] ; store sse
+ sub r0, r11, r0, asr #8 ; return (sse - ((sum * sum) >> 8))
+
+ ldmfd sp!, {r4-r12, pc}
+
+ ENDP
+
+c80808080
+ DCD 0x80808080
+
+ END
+
diff --git a/vp8/encoder/arm/armv6/vp8_variance_halfpixvar16x16_hv_armv6.asm b/vp8/encoder/arm/armv6/vp8_variance_halfpixvar16x16_hv_armv6.asm
new file mode 100644
index 0000000..f9ae3b7
--- /dev/null
+++ b/vp8/encoder/arm/armv6/vp8_variance_halfpixvar16x16_hv_armv6.asm
@@ -0,0 +1,216 @@
+;
+; Copyright (c) 2011 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+ EXPORT |vp8_variance_halfpixvar16x16_hv_armv6|
+
+ ARM
+ REQUIRE8
+ PRESERVE8
+
+ AREA ||.text||, CODE, READONLY, ALIGN=2
+
+; r0 unsigned char *src_ptr
+; r1 int source_stride
+; r2 unsigned char *ref_ptr
+; r3 int recon_stride
+; stack unsigned int *sse
+|vp8_variance_halfpixvar16x16_hv_armv6| PROC
+
+ stmfd sp!, {r4-r12, lr}
+ mov r8, #0 ; initialize sum = 0
+ ldr r10, c80808080
+ mov r11, #0 ; initialize sse = 0
+ mov r12, #16 ; set loop counter to 16 (=block height)
+ mov lr, #0 ; constant zero
+loop
+ add r9, r0, r1 ; pointer to pixels on the next row
+ ; 1st 4 pixels
+ ldr r4, [r0, #0] ; load source pixels a, row N
+ ldr r6, [r0, #1] ; load source pixels b, row N
+ ldr r5, [r9, #0] ; load source pixels c, row N+1
+ ldr r7, [r9, #1] ; load source pixels d, row N+1
+
+ ; x = (a + b + 1) >> 1, interpolate pixels horizontally on row N
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+ ; y = (c + d + 1) >> 1, interpolate pixels horizontally on row N+1
+ mvn r7, r7
+ uhsub8 r5, r5, r7
+ eor r5, r5, r10
+ ; z = (x + y + 1) >> 1, interpolate half pixel values vertically
+ mvn r5, r5
+ uhsub8 r4, r4, r5
+ ldr r5, [r2, #0] ; load 4 ref pixels
+ eor r4, r4, r10
+
+ usub8 r6, r4, r5 ; calculate difference
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+ ; calculate total sum
+ adds r8, r8, r4 ; add positive differences to sum
+ subs r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 2nd 4 pixels
+ ldr r4, [r0, #4] ; load source pixels a, row N
+ ldr r6, [r0, #5] ; load source pixels b, row N
+ ldr r5, [r9, #4] ; load source pixels c, row N+1
+
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+ ldr r7, [r9, #5] ; load source pixels d, row N+1
+
+ ; x = (a + b + 1) >> 1, interpolate pixels horizontally on row N
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+ ; y = (c + d + 1) >> 1, interpolate pixels horizontally on row N+1
+ mvn r7, r7
+ uhsub8 r5, r5, r7
+ eor r5, r5, r10
+ ; z = (x + y + 1) >> 1, interpolate half pixel values vertically
+ mvn r5, r5
+ uhsub8 r4, r4, r5
+ ldr r5, [r2, #4] ; load 4 ref pixels
+ eor r4, r4, r10
+
+ usub8 r6, r4, r5 ; calculate difference
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 3rd 4 pixels
+ ldr r4, [r0, #8] ; load source pixels a, row N
+ ldr r6, [r0, #9] ; load source pixels b, row N
+ ldr r5, [r9, #8] ; load source pixels c, row N+1
+
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+ ldr r7, [r9, #9] ; load source pixels d, row N+1
+
+ ; x = (a + b + 1) >> 1, interpolate pixels horizontally on row N
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+ ; y = (c + d + 1) >> 1, interpolate pixels horizontally on row N+1
+ mvn r7, r7
+ uhsub8 r5, r5, r7
+ eor r5, r5, r10
+ ; z = (x + y + 1) >> 1, interpolate half pixel values vertically
+ mvn r5, r5
+ uhsub8 r4, r4, r5
+ ldr r5, [r2, #8] ; load 4 ref pixels
+ eor r4, r4, r10
+
+ usub8 r6, r4, r5 ; calculate difference
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 4th 4 pixels
+ ldr r4, [r0, #12] ; load source pixels a, row N
+ ldr r6, [r0, #13] ; load source pixels b, row N
+ ldr r5, [r9, #12] ; load source pixels c, row N+1
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+ ldr r7, [r9, #13] ; load source pixels d, row N+1
+
+ ; x = (a + b + 1) >> 1, interpolate pixels horizontally on row N
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+ ; y = (c + d + 1) >> 1, interpolate pixels horizontally on row N+1
+ mvn r7, r7
+ uhsub8 r5, r5, r7
+ eor r5, r5, r10
+ ; z = (x + y + 1) >> 1, interpolate half pixel values vertically
+ mvn r5, r5
+ uhsub8 r4, r4, r5
+ ldr r5, [r2, #12] ; load 4 ref pixels
+ eor r4, r4, r10
+
+ usub8 r6, r4, r5 ; calculate difference
+ add r0, r0, r1 ; set src_ptr to next row
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ add r2, r2, r3 ; set dst_ptr to next row
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+ subs r12, r12, #1
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+ bne loop
+
+ ; return stuff
+ ldr r6, [sp, #40] ; get address of sse
+ mul r0, r8, r8 ; sum * sum
+ str r11, [r6] ; store sse
+ sub r0, r11, r0, asr #8 ; return (sse - ((sum * sum) >> 8))
+
+ ldmfd sp!, {r4-r12, pc}
+
+ ENDP
+
+c80808080
+ DCD 0x80808080
+
+ END
diff --git a/vp8/encoder/arm/armv6/vp8_variance_halfpixvar16x16_v_armv6.asm b/vp8/encoder/arm/armv6/vp8_variance_halfpixvar16x16_v_armv6.asm
new file mode 100644
index 0000000..9e0a035
--- /dev/null
+++ b/vp8/encoder/arm/armv6/vp8_variance_halfpixvar16x16_v_armv6.asm
@@ -0,0 +1,178 @@
+;
+; Copyright (c) 2011 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+ EXPORT |vp8_variance_halfpixvar16x16_v_armv6|
+
+ ARM
+ REQUIRE8
+ PRESERVE8
+
+ AREA ||.text||, CODE, READONLY, ALIGN=2
+
+; r0 unsigned char *src_ptr
+; r1 int source_stride
+; r2 unsigned char *ref_ptr
+; r3 int recon_stride
+; stack unsigned int *sse
+|vp8_variance_halfpixvar16x16_v_armv6| PROC
+
+ stmfd sp!, {r4-r12, lr}
+ mov r8, #0 ; initialize sum = 0
+ ldr r10, c80808080
+ mov r11, #0 ; initialize sse = 0
+ mov r12, #16 ; set loop counter to 16 (=block height)
+ mov lr, #0 ; constant zero
+loop
+ add r9, r0, r1 ; set src pointer to next row
+ ; 1st 4 pixels
+ ldr r4, [r0, #0] ; load 4 src pixels
+ ldr r6, [r9, #0] ; load 4 src pixels from next row
+ ldr r5, [r2, #0] ; load 4 ref pixels
+
+ ; bilinear interpolation
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+
+ usub8 r6, r4, r5 ; calculate difference
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+ ; calculate total sum
+ adds r8, r8, r4 ; add positive differences to sum
+ subs r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 2nd 4 pixels
+ ldr r4, [r0, #4] ; load 4 src pixels
+ ldr r6, [r9, #4] ; load 4 src pixels from next row
+ ldr r5, [r2, #4] ; load 4 ref pixels
+
+ ; bilinear interpolation
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+ usub8 r6, r4, r5 ; calculate difference
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 3rd 4 pixels
+ ldr r4, [r0, #8] ; load 4 src pixels
+ ldr r6, [r9, #8] ; load 4 src pixels from next row
+ ldr r5, [r2, #8] ; load 4 ref pixels
+
+ ; bilinear interpolation
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+ usub8 r6, r4, r5 ; calculate difference
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 4th 4 pixels
+ ldr r4, [r0, #12] ; load 4 src pixels
+ ldr r6, [r9, #12] ; load 4 src pixels from next row
+ ldr r5, [r2, #12] ; load 4 ref pixels
+
+ ; bilinear interpolation
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+ usub8 r6, r4, r5 ; calculate difference
+ add r0, r0, r1 ; set src_ptr to next row
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ add r2, r2, r3 ; set dst_ptr to next row
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+
+ subs r12, r12, #1
+
+ bne loop
+
+ ; return stuff
+ ldr r6, [sp, #40] ; get address of sse
+ mul r0, r8, r8 ; sum * sum
+ str r11, [r6] ; store sse
+ sub r0, r11, r0, asr #8 ; return (sse - ((sum * sum) >> 8))
+
+ ldmfd sp!, {r4-r12, pc}
+
+ ENDP
+
+c80808080
+ DCD 0x80808080
+
+ END
+
diff --git a/vp8/encoder/arm/encodemb_arm.c b/vp8/encoder/arm/encodemb_arm.c
deleted file mode 100644
index 88ad3fc..0000000
--- a/vp8/encoder/arm/encodemb_arm.c
+++ /dev/null
@@ -1,31 +0,0 @@
-/*
- * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
- *
- * Use of this source code is governed by a BSD-style license
- * that can be found in the LICENSE file in the root of the source
- * tree. An additional intellectual property rights grant can be found
- * in the file PATENTS. All contributing project authors may
- * be found in the AUTHORS file in the root of the source tree.
- */
-
-
-#include "vp8/encoder/encodemb.h"
-#include "vp8/common/reconinter.h"
-#include "vp8/encoder/quantize.h"
-#include "vp8/common/invtrans.h"
-#include "vp8/common/recon.h"
-#include "vp8/common/reconintra.h"
-#include "vp8/encoder/dct.h"
-#include "vpx_mem/vpx_mem.h"
-
-extern void vp8_subtract_b_neon_func(short *diff, unsigned char *src, unsigned char *pred, int stride, int pitch);
-
-void vp8_subtract_b_neon(BLOCK *be, BLOCKD *bd, int pitch)
-{
- unsigned char *src_ptr = (*(be->base_src) + be->src);
- short *diff_ptr = be->src_diff;
- unsigned char *pred_ptr = bd->predictor;
- int src_stride = be->src_stride;
-
- vp8_subtract_b_neon_func(diff_ptr, src_ptr, pred_ptr, src_stride, pitch);
-}
diff --git a/vp8/encoder/arm/encodemb_arm.h b/vp8/encoder/arm/encodemb_arm.h
index 8fe4537..bf417fe 100644
--- a/vp8/encoder/arm/encodemb_arm.h
+++ b/vp8/encoder/arm/encodemb_arm.h
@@ -12,6 +12,24 @@
#ifndef ENCODEMB_ARM_H
#define ENCODEMB_ARM_H
+#if HAVE_ARMV6
+extern prototype_subb(vp8_subtract_b_armv6);
+extern prototype_submby(vp8_subtract_mby_armv6);
+extern prototype_submbuv(vp8_subtract_mbuv_armv6);
+
+#if !CONFIG_RUNTIME_CPU_DETECT
+#undef vp8_encodemb_subb
+#define vp8_encodemb_subb vp8_subtract_b_armv6
+
+#undef vp8_encodemb_submby
+#define vp8_encodemb_submby vp8_subtract_mby_armv6
+
+#undef vp8_encodemb_submbuv
+#define vp8_encodemb_submbuv vp8_subtract_mbuv_armv6
+#endif
+
+#endif /* HAVE_ARMV6 */
+
#if HAVE_ARMV7
//extern prototype_berr(vp8_block_error_c);
//extern prototype_mberr(vp8_mbblock_error_c);
diff --git a/vp8/encoder/arm/neon/subtract_neon.asm b/vp8/encoder/arm/neon/subtract_neon.asm
index 3ea00f8..68c2950 100644
--- a/vp8/encoder/arm/neon/subtract_neon.asm
+++ b/vp8/encoder/arm/neon/subtract_neon.asm
@@ -8,45 +8,58 @@
; be found in the AUTHORS file in the root of the source tree.
;
-
- EXPORT |vp8_subtract_b_neon_func|
+ EXPORT |vp8_subtract_b_neon|
EXPORT |vp8_subtract_mby_neon|
EXPORT |vp8_subtract_mbuv_neon|
+ INCLUDE asm_enc_offsets.asm
+
ARM
REQUIRE8
PRESERVE8
AREA ||.text||, CODE, READONLY, ALIGN=2
-;=========================================
-;void vp8_subtract_b_neon_func(short *diff, unsigned char *src, unsigned char *pred, int stride, int pitch);
-|vp8_subtract_b_neon_func| PROC
- ldr r12, [sp] ;load pitch
- vld1.8 {d0}, [r1], r3 ;load src
- vld1.8 {d1}, [r2], r12 ;load pred
- vld1.8 {d2}, [r1], r3
- vld1.8 {d3}, [r2], r12
- vld1.8 {d4}, [r1], r3
- vld1.8 {d5}, [r2], r12
- vld1.8 {d6}, [r1], r3
- vld1.8 {d7}, [r2], r12
+;void vp8_subtract_b_neon(BLOCK *be, BLOCKD *bd, int pitch)
+|vp8_subtract_b_neon| PROC
+
+ stmfd sp!, {r4-r7}
+
+ ldr r3, [r0, #vp8_block_base_src]
+ ldr r4, [r0, #vp8_block_src]
+ ldr r5, [r0, #vp8_block_src_diff]
+ ldr r3, [r3]
+ ldr r6, [r0, #vp8_block_src_stride]
+ add r3, r3, r4 ; src = *base_src + src
+ ldr r7, [r1, #vp8_blockd_predictor]
+
+ vld1.8 {d0}, [r3], r6 ;load src
+ vld1.8 {d1}, [r7], r2 ;load pred
+ vld1.8 {d2}, [r3], r6
+ vld1.8 {d3}, [r7], r2
+ vld1.8 {d4}, [r3], r6
+ vld1.8 {d5}, [r7], r2
+ vld1.8 {d6}, [r3], r6
+ vld1.8 {d7}, [r7], r2
vsubl.u8 q10, d0, d1
vsubl.u8 q11, d2, d3
vsubl.u8 q12, d4, d5
vsubl.u8 q13, d6, d7
- mov r12, r12, lsl #1
+ mov r2, r2, lsl #1
- vst1.16 {d20}, [r0], r12 ;store diff
- vst1.16 {d22}, [r0], r12
- vst1.16 {d24}, [r0], r12
- vst1.16 {d26}, [r0], r12
+ vst1.16 {d20}, [r5], r2 ;store diff
+ vst1.16 {d22}, [r5], r2
+ vst1.16 {d24}, [r5], r2
+ vst1.16 {d26}, [r5], r2
+ ldmfd sp!, {r4-r7}
bx lr
+
ENDP
+
;==========================================
;void vp8_subtract_mby_neon(short *diff, unsigned char *src, unsigned char *pred, int stride)
|vp8_subtract_mby_neon| PROC
diff --git a/vp8/encoder/arm/variance_arm.c b/vp8/encoder/arm/variance_arm.c
index ed1fb16..e77be9f 100644
--- a/vp8/encoder/arm/variance_arm.c
+++ b/vp8/encoder/arm/variance_arm.c
@@ -57,51 +57,38 @@
unsigned short first_pass[36*16];
unsigned char second_pass[20*16];
const short *HFilter, *VFilter;
+ unsigned int var;
- HFilter = vp8_bilinear_filters[xoffset];
- VFilter = vp8_bilinear_filters[yoffset];
+ if (xoffset == 4 && yoffset == 0)
+ {
+ var = vp8_variance_halfpixvar16x16_h_armv6(src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, sse);
+ }
+ else if (xoffset == 0 && yoffset == 4)
+ {
+ var = vp8_variance_halfpixvar16x16_v_armv6(src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, sse);
+ }
+ else if (xoffset == 4 && yoffset == 4)
+ {
+ var = vp8_variance_halfpixvar16x16_hv_armv6(src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, sse);
+ }
+ else
+ {
+ HFilter = vp8_bilinear_filters[xoffset];
+ VFilter = vp8_bilinear_filters[yoffset];
- vp8_filter_block2d_bil_first_pass_armv6(src_ptr, first_pass,
- src_pixels_per_line,
- 17, 16, HFilter);
- vp8_filter_block2d_bil_second_pass_armv6(first_pass, second_pass,
- 16, 16, 16, VFilter);
+ vp8_filter_block2d_bil_first_pass_armv6(src_ptr, first_pass,
+ src_pixels_per_line,
+ 17, 16, HFilter);
+ vp8_filter_block2d_bil_second_pass_armv6(first_pass, second_pass,
+ 16, 16, 16, VFilter);
- return vp8_variance16x16_armv6(second_pass, 16, dst_ptr,
- dst_pixels_per_line, sse);
-}
-
-unsigned int vp8_variance_halfpixvar16x16_h_armv6(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{
- return vp8_sub_pixel_variance16x16_armv6(src_ptr, source_stride, 4, 0,
- ref_ptr, recon_stride, sse);
-}
-
-unsigned int vp8_variance_halfpixvar16x16_v_armv6(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{
- return vp8_sub_pixel_variance16x16_armv6(src_ptr, source_stride, 0, 4,
- ref_ptr, recon_stride, sse);
-}
-
-unsigned int vp8_variance_halfpixvar16x16_hv_armv6(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{
- return vp8_sub_pixel_variance16x16_armv6(src_ptr, source_stride, 4, 4,
- ref_ptr, recon_stride, sse);
+ var = vp8_variance16x16_armv6(second_pass, 16, dst_ptr,
+ dst_pixels_per_line, sse);
+ }
+ return var;
}
#endif /* HAVE_ARMV6 */
diff --git a/vp8/encoder/asm_enc_offsets.c b/vp8/encoder/asm_enc_offsets.c
index c7983c1..9c81c8d 100644
--- a/vp8/encoder/asm_enc_offsets.c
+++ b/vp8/encoder/asm_enc_offsets.c
@@ -48,6 +48,14 @@
DEFINE(vp8_blockd_dqcoeff, offsetof(BLOCKD, dqcoeff));
DEFINE(vp8_blockd_eob, offsetof(BLOCKD, eob));
+// subtract
+DEFINE(vp8_block_base_src, offsetof(BLOCK, base_src));
+DEFINE(vp8_block_src, offsetof(BLOCK, src));
+DEFINE(vp8_block_src_diff, offsetof(BLOCK, src_diff));
+DEFINE(vp8_block_src_stride, offsetof(BLOCK, src_stride));
+
+DEFINE(vp8_blockd_predictor, offsetof(BLOCKD, predictor));
+
//pack tokens
DEFINE(vp8_writer_lowvalue, offsetof(vp8_writer, lowvalue));
DEFINE(vp8_writer_range, offsetof(vp8_writer, range));
diff --git a/vp8/encoder/onyx_if.c b/vp8/encoder/onyx_if.c
index e5d2bd8..a18447d 100644
--- a/vp8/encoder/onyx_if.c
+++ b/vp8/encoder/onyx_if.c
@@ -3258,7 +3258,8 @@
}
#if CONFIG_MULTITHREAD
- sem_post(&cpi->h_event_end_lpf); /* signal that we have set filter_level */
+ if (cpi->b_multi_threaded)
+ sem_post(&cpi->h_event_end_lpf); /* signal that we have set filter_level */
#endif
if (cm->filter_level > 0)
diff --git a/vp8/vp8cx_arm.mk b/vp8/vp8cx_arm.mk
index 349c3fd..165dada 100644
--- a/vp8/vp8cx_arm.mk
+++ b/vp8/vp8cx_arm.mk
@@ -15,7 +15,6 @@
# encoder
VP8_CX_SRCS-$(ARCH_ARM) += encoder/arm/arm_csystemdependent.c
-VP8_CX_SRCS-$(HAVE_ARMV7) += encoder/arm/encodemb_arm.c
VP8_CX_SRCS-$(HAVE_ARMV7) += encoder/arm/quantize_arm.c
VP8_CX_SRCS-$(HAVE_ARMV7) += encoder/arm/picklpf_arm.c
VP8_CX_SRCS-$(HAVE_ARMV6) += encoder/arm/dct_arm.c
@@ -34,10 +33,14 @@
#File list for armv6
# encoder
+VP8_CX_SRCS-$(HAVE_ARMV6) += encoder/arm/armv6/vp8_subtract_armv6$(ASM)
VP8_CX_SRCS-$(HAVE_ARMV6) += encoder/arm/armv6/vp8_fast_fdct4x4_armv6$(ASM)
VP8_CX_SRCS-$(HAVE_ARMV6) += encoder/arm/armv6/vp8_fast_quantize_b_armv6$(ASM)
VP8_CX_SRCS-$(HAVE_ARMV6) += encoder/arm/armv6/vp8_sad16x16_armv6$(ASM)
VP8_CX_SRCS-$(HAVE_ARMV6) += encoder/arm/armv6/vp8_variance16x16_armv6$(ASM)
+VP8_CX_SRCS-$(HAVE_ARMV6) += encoder/arm/armv6/vp8_variance_halfpixvar16x16_h_armv6$(ASM)
+VP8_CX_SRCS-$(HAVE_ARMV6) += encoder/arm/armv6/vp8_variance_halfpixvar16x16_v_armv6$(ASM)
+VP8_CX_SRCS-$(HAVE_ARMV6) += encoder/arm/armv6/vp8_variance_halfpixvar16x16_hv_armv6$(ASM)
VP8_CX_SRCS-$(HAVE_ARMV6) += encoder/arm/armv6/vp8_mse16x16_armv6$(ASM)
VP8_CX_SRCS-$(HAVE_ARMV6) += encoder/arm/armv6/vp8_variance8x8_armv6$(ASM)
VP8_CX_SRCS-$(HAVE_ARMV6) += encoder/arm/armv6/walsh_v6$(ASM)
diff --git a/vpxenc.c b/vpxenc.c
index 6c13cd1..39256b6 100755
--- a/vpxenc.c
+++ b/vpxenc.c
@@ -236,7 +236,13 @@
stats->buf_ptr = new_ptr + (stats->buf_ptr - (char *)stats->buf.buf);
stats->buf.buf = new_ptr;
stats->buf_alloc_sz = new_sz;
- } /* else ... */
+ }
+ else
+ {
+ fprintf(stderr,
+ "\nFailed to realloc firstpass stats buffer.\n");
+ exit(EXIT_FAILURE);
+ }
}
memcpy(stats->buf_ptr, pkt, len);
@@ -698,10 +704,18 @@
/* Save a cue point if this is a keyframe. */
if(is_keyframe)
{
- struct cue_entry *cue;
+ struct cue_entry *cue, *new_cue_list;
- glob->cue_list = realloc(glob->cue_list,
- (glob->cues+1) * sizeof(struct cue_entry));
+ new_cue_list = realloc(glob->cue_list,
+ (glob->cues+1) * sizeof(struct cue_entry));
+ if(new_cue_list)
+ glob->cue_list = new_cue_list;
+ else
+ {
+ fprintf(stderr, "\nFailed to realloc cue list.\n");
+ exit(EXIT_FAILURE);
+ }
+
cue = &glob->cue_list[glob->cues];
cue->time = glob->cluster_timecode;
cue->loc = glob->cluster_pos;
